DIA, DDA and PRM strategies for biomarker identification and validation in plasma cohorts of COVID-19 patients with varying severity

High-throughput Mass Spectrometry (MS)-based proteomics has generated extensive data on COVID-19, enhancing our understanding of the disease. Studies on COVID-19 patients show significant impact of the infection on plasma and sera proteome [1–3]. These findings have spurred research into biomarkers for diagnosis, prognosis or therapy [4,5]. Improved MS methodologies for accurate proteome profiling and evaluation promise better quantity and quality of data, as well as the cost-benefit quantification [6]. These methods include label free approaches, based on Data-Independent Acquisition (DIA) or Data-Dependent Acquisition (DDA) strategies, to scan for the whole proteome or the Parallel Reaction Monitoring (PRM) strategy, for validating known proteins without needing specific antibodies.

Despite extensive use of DIA, DDA and PRM in studying diseases like COVID-19 [7,8], the concordance between these technologies for biomarker identification has not been formally evaluated in clinical human samples.

This study aimed to evaluate the concordance of these three MS proteomics approaches in identifying and assessing biomarkers discovery in the plasma of COVID-19 patients with varying grades of severity. We focused on how DIA-based strategy enhances candidate identification versus DDA and compared protein quantification reliability between DIA, DDA and PRM. Concordance between DIA and DDA were conducted on samples from 48 hospitalized and 48 non-hospitalized patients (N = 96). Among them, samples from 10 patients in each group were available for DIA-PRM comparisons (63 proteins).

PRM data showed higher statistical power and better biomarker identification than DIA, which in turn was superior to DDA. All technologies exhibited excellent concordance in Fold-Change and good sample-level correlation. However, systematic differences between PRM and DIA, and between DIA and DDA, highlight the need for calibration to make their quantifications comparable. We achieved this with simple linear regression, which performed outstandingly.

Finally, we analyzed by PRM a cohort of 170 individuals and we selected 10 proteins as clinically relevant for further validation as infection prognosis biomarkers.

References

Shen, B. et al. Cell 182, 59–72.e15 (2020).Messner, C. B. et al. Cell Syst 11, 11–24.e4 (2020).Whetton, A. D. et al. J. Proteome Res. 19, 4219–4232 (2020).Bouhaddou, M. et al. Cell 182, 685–712.e19 (2020).Bojkova, D. et al. Nature 583, 469–472 (2020).Bantscheff, M. et al. Anal. Bioanal. Chem. 404, 939–965 (2012).Geyer, P. E. et al. EMBO Mol. Med. 13, e14167 (2021).Demichev, V. et al. Cell Syst 12, 780–794.e7 (2021).